Shlomo Pongratz reported GRO L2 header check was suited for Ethernet
only, and failed on IB/ipoib traffic.

He provided a patch faking a zeroed header to let GRO aggregates frames.

Roland Dreier, Herbert Xu, and others suggested we change GRO L2 header
check to be more generic, ie not assuming L2 header is 14 bytes, but
taking into account hard_header_len.

__napi_gro_receive() has special handling for the common case (Ethernet)
to avoid a memcmp() call and use an inline optimized function instead.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Reported-by: NShlomo Pongratz <shlomop@mellanox.com>
Cc: Roland Dreier <roland@kernel.org>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Tested-by: NSean Hefty <sean.hefty@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit a0417fa3 ("net: Make qdisc_skb_cb upper size bound
explicit.") made it possible for a netdev driver to use skb->cb
between its header_ops.create method and its .ndo_start_xmit
method.  Use this in ipoib_hard_header() to stash away the LL address
(GID + QPN), instead of the "ipoib_pseudoheader" hack.  This allows
IPoIB to stop lying about its hard_header_len, which will let us fix
the L2 check for GRO.
Signed-off-by: NRoland Dreier <roland@purestorage.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

8<----------------------------------------------------------------------
From: Ralf Roesch <ralf.roesch@rw-gmbh.de>
Date: Wed, 16 Nov 2011 09:33:50 +0100
Subject: [PATCH] net: enable TC35815 for MIPS again

TX493[8,9] MIPS SoCs support 2 Ethernet channels of type TC35815
which are connected to the internal PCI controller.
And JMR3927 MIPS board has a TC35815 chip on board.
These dependencies were lost on movement to drivers/net/ethernet/toshiba.
Signed-off-by: NRalf Roesch <ralf.roesch@rw-gmbh.de>
Signed-off-by: NAtsushi Nemoto <anemo@mba.ocn.ne.jp>
CC: stable@vger.kernel.org [3.2+]
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When this GETHER controller received a large frame (about 1800 bytes
or more), skb_over_panic() happened. This is because the previous
driver set the RFLR to 0x1000 (4096 bytes) and the skb allocate size
is smaller than 4096 bytes. So, the controller accepted such a frame.

The controller can discard a large frame by the RFLR setting.
So, the patch modifies the value of RFLR to mtu + ETH_HLEN +
VLAN_HLEN + ETH_FCS_LEN.
Signed-off-by: NYoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Tested-by: NAnisse Astier <anisse@astier.eu>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We allocate memory for 'new_data' with kmalloc(). If we get the memory
we then try to build_skb() and if that should fail (which it can) we
do not enter 'if (likely(skb)) {' and actually use 'new_data' but
instead fall through to the 'drop:' label and end up returning from
the function without ever assigning 'new'data' to anything or freeing
it. That leaks the memory allocated to 'new_data'.

This patch fixes the memory leak by doing a kfree(new_data) in the
case where build_skb() fails (or where allocation of 'new_data' itself
fails, but in taht case it's just a harmless kfree(NULL)).
Signed-off-by: NJesper Juhl <jj@chaosbits.net>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Acked-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As it is, with PCI/ISA/MCA/CCW all set to n and PCMCIA set to m
setting TR to y will set LLC to m, with very unpleasant results -
net/802/psnap gets picked into obj-y, resulting in the kernel
that won't link - psnap calls functions from llc.  The cause,
AFAICS, is that kconfig gets rev_dep for LLC containing
|| TR && (deps for TR)
and even though TR is boolean, both LLC and PCMCIA are tristate
and that thing becomes || y && (n || m), i.e. || m.  The reason
for dependency on PCMCIA is that when none of PCI, ISA, MCA, CCW
or PCMCIA is set there'll be no tokenring drivers, so there's no
point building tokenring core.  Proper fix probably belongs in
kconfig (we need strict and, such that y <strict_and> m would be
y, so that rev_deps added for tristate selected by bool would
use that instead of &&; we'd have || TR <strict_and> (deps for TR)
in this case), but it's a rather intrusive change.  There's an
easy workaround in case of TR -> LLC select, namely to have a def_bool y
symbol sitting under if TR and have that symbol selecting LLC.
Kudos to johill for suggesting that one...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As reported by several people...

The code in rx_clean was panic'ing so revert
commit d0249e44.
Will redo DMA mapping checks as new patches for a later release.
Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

mlx4_en_prepare_rx_desc does not correctly clean up after it finds an
allocation failure. It should unmap a page before calling put_page, but
it only calls the later.

This bug would prevent a device removal using hotplug after setting the
device MTU to 9000 and opening the network interface. After the fix, we
still see the allocation failure with MTU 9000, but we are able to
remove the device.
Signed-off-by: NThadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After opening the network interface, Mellanox ConnectX device cannot be
removed by hotplug because it has not properly unmapped all DMA memory.

It happens that mlx4_en_activate_rx_rings overrides the variable that
keeps the size of the memory mapped.

This is fixed by passing to mlx4_en_destroy_rx_ring the same size that is
given to mlx4_en_create_rx_ring.

After applying this patch, hot unplugging the device works after opening
the interface.
Signed-off-by: NThadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use port number for correct steering (list per port).
Before the fix all steering entries (for both physical ports)
were managed in first port structures, so we had leakage of resources
for port 2.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use MLX4_DEV_CAP_FLAG_VEP_UC_STEER for unicast_promisc_add/remove
Unicast entries were managed in wrong data structures.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Perform cleanup also in non-master flow.
The VFs use communication channel as well.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.co.il>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Initially diagnosed on Ubuntu 11.04 with kernel 2.6.38.

velocity_close is not called during a suspend / resume cycle in this
driver and it has no business playing directly with power states.
Signed-off-by: NDavid Lv <DavidLv@viatech.com.cn>
Acked-by: NFrancois Romieu <romieu@fr.zoreil.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Eric Dumazet found that commit 813b3b5d
(ipv4: Use caller's on-stack flowi as-is in output
route lookups.) that comes in 3.0 added a regression.
The problem appears to be that resulting flowi4_oif is
used incorrectly as input parameter to some routing lookups.
The result is that when connecting to local port without
listener if the IP address that is used is not on a loopback
interface we incorrectly assign RTN_UNICAST to the output
route because no route is matched by oif=lo. The RST packet
can not be sent immediately by tcp_v4_send_reset because
it expects RTN_LOCAL.

	So, change ip_route_connect and ip_route_newports to
update the flowi4 fields that are input parameters because
we do not want unnecessary binding to oif.

	To make it clear what are the input parameters that
can be modified during lookup and to show which fields of
floiw4 are reused add a new function to update the flowi4
structure: flowi4_update_output.

Thanks to Yurij M. Plotnikov for providing a bug report including a
program to reproduce the problem.

Thanks to Eric Dumazet for tracking the problem down to
tcp_v4_send_reset and providing initial fix.
Reported-by: NYurij M. Plotnikov <Yurij.Plotnikov@oktetlabs.ru>
Signed-off-by: NJulian Anastasov <ja@ssi.bg>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Binding RST packet outgoing interface to incoming interface
for tcp v4 when there is no socket associate with it.
when sk is not NULL, using sk->sk_bound_dev_if instead.
(suggested by Eric Dumazet).

This has few benefits:
1. tcp_v6_send_reset already did that.
2. This helps tcp connect with SO_BINDTODEVICE set. When
connection is lost, we still able to sending out RST using
same interface.
3. we are sending reply, it is most likely to be succeed
if iif is used
Signed-off-by: NShawn Lu <shawn.lu@ericsson.com>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It was recently pointed out to me that the get_prioidx function sets a bit in
the prioidx map prior to checking to see if the index being set is out of
bounds.  This patch corrects that, avoiding the possiblity of us writing beyond
the end of the array
Signed-off-by: NNeil Horman <nhorman@tuxdriver.com>
Reported-by: NStanislaw Gruszka <sgruszka@redhat.com>
CC: Stanislaw Gruszka <sgruszka@redhat.com>
CC: "David S. Miller" <davem@davemloft.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless into for-davem

Correct spelling "packtetbuf_num" to "packetbuf_num" in
drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
Signed-off-by: NMasanari Iida <standby24x7@gmail.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>

Change copyright date to 2012 in the source files.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Tested-by: NRobert E Garrett <robertX.e.garrett@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NGreg Rose <gregory.v.rose@intel.com>
Tested-by: NRobert E Garrett <robertX.e.garrett@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Remove jump to out label that was skipping over code to replace the
ack bit in the mailbox cache variable.  This was causing driver mailbox
routines that poll for the PF to ack a message to time out which would
in turn cause all other mailbox messages to fail.  Also replace the
mailbox cache variable message indication bit when a message is found
so that other functions that choose to poll for a mailbox message from
the PF won't miss it.
Signed-off-by: NGreg Rose <gregory.v.rose@intel.com>
Tested-by: NRobert E Garrett <robertX.e.garrett@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NStephen Hemminger <shemminger@vyatta.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Send message to system log when the VF requests another MAC filter
but there is no space left on the device for it.
Signed-off-by: NGreg Rose <gregory.v.rose@intel.com>
Tested-by: NRobert E Garrett <robertX.e.garrett@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

New year so bump the copyright date.
Signed-off-by: NDon Skidmore <donald.c.skidmore@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Change the way channels objects are linked together by peak_pci_probe()
avoiding any kernel oops when driver is removed. Side effect is that
the list is now browsed from last to first channel.
Signed-off-by: NStephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

netif_device_attched is called twice when ems_usb_start fails with -ENODEV
Signed-off-by: NSebastian Haas <dev@sebastianhaas.info>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

The test (((errc & PCH_REC) >> 8) > 127) would always be false because
the receive error counter ((errc & PCH_REC) >> 8) is at most 127, where
PCH_REC is defined as 0x7f00.  To test whether the receive error counter
has reached the error passive level, the RP bit (15) should be used.
Signed-off-by: NXi Wang <xi.wang@gmail.com>
Acked-by: NWolfgang Grandegger <wg@grandegger.com>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

This patch fixes locking problems by using netif_rx() instead of
netif_receive_skb() in ti_hecc_error() called from the interrupt
context.
Signed-off-by: NWolfgang Grandegger <wg@grandegger.com>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

This fix avoids a deadlock if an interrupt occurs
during consecutive port operations on ISA cards
utilising indirect access via address and data
port.

Tested on a B&R ISA card.

Cc: linux-can@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: NWolfgang Zarre <lkdev@essax.com>
Acked-by: NWolfgang Grandegger <wg@grandegger.com>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

On slow systems and high CAN bitrates, the error message
"can_put_echo_skb: BUG! echo_skb is occupied!" did show up because
can_put_echo_skb() was called after starting the transfer.
Signed-off-by: NWolfgang Grandegger <wg@grandegger.com>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

As pointed out by Reuben Dowle and Lothar Waßmann, the TWRN_INT,
RWRN_INT, BOFF_INT interrupt sources need to be cleared as well
to avoid interrupt flooding, at least for the Flexcan on i.MX28
SOCs. Furthermore, the interrupts are only cleared, if really one
of those interrupt sources are pending (which is not the case for
rx and tx done).

Cc: Reuben Dowle <Reuben.Dowle@navico.com>
Cc: Lothar Waßmann <LW@KARO-electronics.de>
Signed-off-by: NWolfgang Grandegger <wg@grandegger.com>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>

SKB is freed twice upon send error. The Network stack consumes SKB even
when it returns error code.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Always use cfmuxl_remove_uplayer when removing a up-layer.
cfmuxl_ctrlcmd() can be called independently and in parallel with
cfmuxl_remove_uplayer(). The race between them could cause list_del_rcu
to be called on a node which has been already taken out from the list.
That lead to a (rare) crash on accessing poisoned node->prev inside
list_del_rcu.

This fix ensures that deletion are done holding the same lock.
Reported-by: NDmitry Tarnyagin <dmitry.tarnyagin@stericsson.com>
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is a possible data corruption if an RNDIS message goes beyond page
boundary in the sending code path. This patch fixes the problem.
Signed-off-by: NHaiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: NK. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For code path not on the xmit, use netif_tx_disable() instead of
netif_stop_queue() to ensure other CPUs are not doing xmit.
Signed-off-by: NHaiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: NK. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

PTR_ERR should be called before its argument is cleared.

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression e,e1;
constant c;
@@

*e = c
... when != e = e1
    when != &e
    when != true IS_ERR(e)
*PTR_ERR(e)
// </smpl>
Signed-off-by: NJulia Lawall <Julia.Lawall@lip6.fr>
Reported-by: NJosh Triplett <josh@joshtriplett.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The packets with size larger than 1452 will be dropped by bridge
which with two hyperv netdevice ports. This cause by hyperv netvsc
driver always copy the trailer padding to the data packet, and then
the skb received from netdevice may include wrong skb->len (20 bytes
larger than the real size normally). The captured packet may like
this:

  Ethernet II, Src: Microsof_00:00:07 (00:15:5d:00:00:07),
               Dst: HewlettP_00:00:4e (00:1f:29:00:00:4e)
    Destination: HewlettP_e6:00:4e (00:1f:29:00:00:4e)
    Source: Microsof_f6:6d:07 (00:15:5d:f6:6d:07)
    Type: IP (0x0800)
    Trailer: 1415161718191A1B1C1D1E1F20212223
    Frame check sequence: 0x24252627 [incorrect, should be 0x7c2e5a5e]

The following command help to reproduction it, and the ping ICMP
packets will be dropped by bridge.
  $ ping ip -s 1453

This patch fixed it by removing the trailer padding from the data
packet.
Signed-off-by: NWei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 4acb4190 tries to fix the using uninitialized value
introduced by commit 3dc43e3e,  but it would make the
per-socket memory limits too small.

This patch fixes this and also remove the redundant codes
introduced in 4acb4190.
Signed-off-by: NJason Wang <jasowang@redhat.com>
Acked-by: NGlauber Costa <glommer@parallels.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>